MICDE funds wide-ranging computational discovery in galactic formation, drug discovery, bacterial biofilm colonies and turbulence simulations

By | News, Research

Since 2017 the Michigan Institute for Computational Discovery & Engineering (MICDE) Catalyst Grants program has funded a wide spectrum of cutting-edge research that combines science, engineering, mathematics and computer science. This year the program will fund four new projects that continue this tradition: Prof. Aaron Frank (Chemistry) and his group will spearhead efficient strategies to rapidly develop treatments for emerging diseases– a need made more compelling by the current COVID-19 Pandemic. Their approach combines generative artificial intelligence models and molecular docking to rapidly explore the space of chemical structures and generate target-specific virtual libraries for drug discovery. Prof. Marisa Eisenberg (Epidemiology, Mathematics, and Complex Systems) and Prof. Alexander Rickard’s (Epidemiology) groups will develop novel computational techniques to study biofilm architectures.  Biofilms are complex assemblages of microbial cells that form on almost any natural and man-made surface. They cause several debilitating diseases, and can even damage machinery and equipment, elevating the understanding of their behaviour to a critical need. Prof. Oleg Gnedin (Astronomy) will develop novel techniques to tailor the mathematical initial conditions from which to simulate chosen regions of the universe. The resulting insights will help uncover the origins of our own galaxy, the Milky Way. Finally, Prof. Aaron Towne (Mechanical Engineering) will advance the modeling of complex, turbulent flows and other large-scale systems in engineering science. His research will enable orders of magnitude of acceleration in the computation of extremely large scale flows in a number of engineering systems.

“These four projects have the potential to catalyze and  reorient the directions of their research fields by developing and harnessing powerful paradigms of computational science”, said Krishna Garikipati, Professor of Mechanical Engineering and of Mathematics, and MICDE’s Director. “MICDE’s mission is to lead the advances in computational science research by bringing together interdisciplinary teams at U of M, and these projects embody that vision.” 

More about MICDE’s catalyst grant program and the projects can be found at micde.umich.edu/catalyst.

Fabricio Vasselai wins the Irving Louis Horowitz Award from the Horowitz Foundation for Social Policy

By | News

Fabricio Vasselai, a dual Ph. D. candidate in Political Science and Scientific Computing is a recipient of this year’s Horowitz Foundation awards from the Horowitz Foundation for Social Policy. His proposal titled “Elections in the AI era: using Machine Learning and Multi-Agent Systems to detect and study menaces to election integrity” won the Irving Louis Horowitz Award, given to the overall most outstanding project of the year, as well as the Joshua Feigenbaum Award as the most outstanding project on Arts, Popular Culture and Mass Communication.

The proposal develops Artificial Intelligence tools to detect and to study threats to election integrity. First, novel Multi-agent simulations of elections (MASE) are derived and implemented to be the data-generating process of synthetic data. Then these data is used to train Supervised Machine Learning (SML) to detect fraud in real election result counts. He uses such ability to create simulated training data to properly bootstrap the SML classifications, allowing for the novel estimation of uncertainty around election fraud detection. He also uses MASE to perform virtual experiments on the spread of fake news, showing that biased misinformation is critical for political polarization to flourish in majoritarian elections.

Fabricio Vasselai is an MICDE Fellow (awarded on 2018), and he is currently a Researcher at U-M’s Center for Political Studies and Center for Complex Systems.

Established in 1998, the Horowitz Foundation awards grants to scholars to assist them in completing their dissertations. It is highly competitive, with less than 3 percent of applicants receiving an award this year.

Microsoft AI for Health Program to support an AI-facilitated Optimization Framework for Improving COVID-19 Testing

By | News, Research

With the recent resurgence of COVID-19 infections, testing has become central to an integrated, global response to the pandemic. Accurate, effective, and efficient testing can lead to early detection and prompt an agile response by public health authorities. Strategic testing systems are critical for providing data that will inform disease prevention, preparation, and intervention. MICDE Associate Director and Associate Professor of Industrial and Operations Engineering and of Civil and Environmental Engineering, Siqian Shen, has recently published an article pin-pointing a number of pivotal operations research and industrial engineering tools that can be brought to  the fight against COVID-19. One of the key lessons from her research is the importance of expanding the availability of COVID-19 testing and making the resulting data transparent to the public as anonymized, summary statistics. This enables informed decision making by individuals, public health officials, and governments.  

Based on these high-impact findings, Professor Shen is striding ahead to design a comprehensive COVID-19 testing framework to efficiently serve the urgent needs of diverse population groups . A grant from Microsoft’s AI for Health program, part of the AI for Good initiative, will provide credits to use Microsoft’s Azure service.  With this cyber resource, Professor Shen and her team will integrate and coordinate decision-making models and data analytics tools that they have developed for testing on a Cloud-based platform. In addition, their AI framework is dynamic, and collects daily infection data to improve testing-related decisions. Such a platform could have significant impacts on three major problems that exist with current testing design strategies:

1) Where to locate testing facilities and how to allocate test kits and other resources.
2) How to effectively triage different population groups through effective appointment scheduling.
3) How to visualize real-time testing capacities to better inform the public and serve ad-hoc needs of patients. 

Prof. Shen’s research will integrate AI techniques with optimization to dynamically refine existing testing design methods for gathering and analyzing data from unexplored populations and regions around the globe. The development and refinement of these new models with the support of Microsoft Azure will create a transparent, data-informed testing system that will allow public health and government authorities to make agile, data-driven decisions to aid in the prevention, preparation, intervention, and management of COVID-19 and other outbreaks of infectious diseases.

Siqian Shen is a  Professor of Industrial and Operations Engineering, and of Civil and Environmental Engineering at the University of Michigan, an Associate Director of the Michigan Institute for Computational Discovery & Engineering, and an affiliated faculty member in the Michigan Institute for Data Science. Her research group works on both theoretical and applied aspects of problems by combining stochastic programming, integer programming, network optimization,  machine learning and statistics.

What is the right model? Different MRIO models yield very different carbon footprints estimates in China

By | Research

Appropriate accounting of greenhouse gas emissions is the first step to assign mitigation responsibilities and develop effective mitigation strategies. Consistent methods are required to fairly assess a region’s impact on climate change. Two leading reasons for the existence of different accounting systems are the political pressures, and the actual costs of climate mitigation to local governments. At the international level there has been consensus, and global environmentally extended multi-regional input-output (EE-MRIO) models that capture the interdependence of and their environmental impacts have been constructed.  However in China, the largest greenhouse gas emitter, where accurate interregional trade-related emission accounts are critical to develop mitigation strategies and monitor progresses at the regional level, this information is sporadic and inconsistent. Prof. Ming Xu from the School of Environment and Sustainability, and his research group, analyzed the available data from China, which dates back to 2012. They showed that the results varied wildly depending on the MRIO model used. For example, they found two MRIO models differed as much as 208 metric tons in a single region, which is equivalent to the emissions of Argentina, United Arab Emirates, or the Netherlands. Their results show the need to prioritize future efforts to harmonize greenhouse gas emissions accounting within China.

Ming Xu is an Associate Professor in the School for Environment and Sustainability and in the Department of Civil and Environmental Engineering at the University of Michigan, Ann Arbor. His research focuses on the broad fields of sustainable engineering and industrial ecology. 

Read the full article.

Modeling the transmission of infectious aerosols

By | Feature, Research

Inhalation of micron-sized droplets represents the dominant transmission mechanism for influenza and rhinovirus, and recent research shows that it is likely also the case for the novel coronavirus.  Increasing evidence suggests that the transmission of infectious aerosols is more complex than previously thought. Coughing, sneezing and even talking yield a gaseous flow field near the infected person that is dynamic and turbulent in nature. Existing models commonly employed in simulations of aerosol transmission attempt to represent the effect of turbulence using random walk models that are often phenomenological in nature, employing adjustable parameters and inherently assuming the turbulent fluctuations ‘felt’ by a droplet do not depend upon direction. To design physics-informed guidelines to minimize the spread of this virus, improved predictive modeling capabilities for effectively tracking the aerosol paths are needed. Dr. Aaron M. Lattanzi and Prof. Jesse Capecelatro, from Mechanical Engineering and MICDE are tackling this problem by focusing on mathematical modeling of aerosol dispersion. They derived analytical solutions for the mean-squared-displacement resulting from systems of stochastic differential equations. A key element of their methodology is that the solution connects stochastic theory inputs to statistics present in high-fidelity simulations or experiments, providing a framework for developing improved models.

Simple simulation of aerosol dispersion from a single-point source. The grey, cone-like surface is the approximation using Force Langevin (FL) theory and the colored particles are from integration of Newton’s equations with stochastic drag forces.

Prof. Capecelatro’s research group develops physics-based models and numerical algorithms to leverage supercomputers for prediction and optimization of the complex flows relevant to energy and the environment. The main focus of their research involves developing robust and scalable numerical tools to investigate the multiphysics and multiscale phenomena under various flow conditions, like those that they study here. They recently submitted their findings to the Journal of Fluid Mechanics, and are continuing to work on this problem hoping it will help understand the transmission of COVID-19 and therefore help optimize current guidelines.

U-M modeling epidemiologists helping navigate the COVID-19 pandemic

By | Feature, News, Research

[top] Screenshoot of the Michigan COVID-19 Modeling Dashboard (epimath.github.io/covid-19-modeling/); [bottom left] Marisa Eisenberg (Epidemiology, Complex Systems and Mathematics); [bottom right] Jonathan Zelner (Epidemiology).

The COVID-19 pandemic is producing massive amounts of information that more often than not lead to different interpretations. The accurate analysis of this daily input of data is crucial to predict possible outcomes and design solutions rapidly. These can only be achieved with expertise in modeling infectious diseases, and with the power of computational science theory and infrastructure. U-M’s Epidemiology Department, in the School of Public Health, has a very strong cohort of researchers who work on mathematically modeling the dynamics of infectious diseases, the analysis of these models, and large scale computer simulations — all to understand the spread and mitigation of pandemics. They are applying their long experience and expertise to the current COVID-19 outbreak, aiding the government make informed decisions, and helping media outlets produce accurate reports for the general public. Marisa Eisenberg, Associate Professor of Epidemiology, of Complex Systems, and of Mathematics, and her colleagues are using a differential equation transmission modeling approach to analyze scenarios and generate short-term forecasts for the COVID-19 epidemic in State of Michigan. They are communicating directly with the Michigan Department of Health and Human Services and providing critical tools, like the Michigan COVID-19 Modeling Dashboard, to inform decision-making. Prof. Eisenberg’s team is helping to forecast the numbers of laboratory-confirmed cases, fatalities, hospitalized patients, and hospital capacity issues (such as ICU beds needed), and examining how social distancing can impact the spread of the epidemic. Prof. Jonathan Zelner, whose research is focused on using spatial and social network analysis and dynamic modeling to prevent infectious diseases, is part of a group helping map the outbreak in Michigan. He also has provided valuable insights to journalists contributing to a better understanding of the situation, including what made New York City so vulnerable to the coronavirus (NYT), the role of wealth inequality during epidemics (CNBC) and what professions and communities are particularly vulnerable to infection (NG). 

Professors Eisenberg and Zelner are not alone in this fight. Many more researchers from U-M’s School of Public Health and throughout campus have risen to the challenges posed by this pandemic. 

U-M Tobacco Center, CAsToR, accepting applications for scholarships to enroll in short U-M summer courses

By | Educational, Funding Opportunities

The Center for the Assessment of Tobacco Regulations (CAsToR) is accepting applications for scholarships to participate in a short course on tobacco simulation modeling, EPID730 Simulation Modeling of Tobacco Use, Health Effects and Policy Impacts,  or in the course EPID 793 Complex Systems Modeling for Public Health Research, to be offered during the University of Michigan Summer Session in Epidemiology (SSE) Program in 2020. Note that the courses will now be offered in an online format only.

New deadline to apply: 11:59 PM EST on Wednesday April 15, 2020

Additional details on both courses can be found here: sph.umich.edu/umsse/courses/1week.html.  A tentative course syllabus for EPID 730 can be found in this google doc. See the full RFA here. Please contact Katie Zarins (kmrents@umich.edu) with questions.

New MOOC in Computational Thinking has launched!

By | Educational, Feature, Happenings

The Michigan Institute for Computational Discovery & Engineering and the University of Michigan Center for Academic Innovation have partnered to launch a Massive Open Online Course (MOOC) titled Problem Solving using Computational Thinking. The idea for this MOOC arose from the team’s recognition of the ubiquity of computation. However, the developers were equally keen to distinguish this offering from MOOCs on programming, and to instead highlight how broader computational thinking also makes its presence felt in somewhat unexpected domains. The MOOC is organized in a series of real-world examples that includes how, using computational thinking, it is possible to help plan and prepare for a flu season, track human rights violations or monitor the safety of crowds. The process of computational thinking that this MOOC focuses on ranges from problem identification, through abstraction to evaluating solutions. Problem Solving using Computational Thinking seeks to introduce students and teachers to the systematic thinking needed to conceptualize a problem with the intent of eventually using some computational tools to solve it.

The developers of this MOOC are drawn from the School of Public Health, the College of Engineering, the School of Education and MICDE. Problem Solving using Computational Thinking is available in Coursera through Michigan Online. To learn more please visit online.umich.edu/courses/problem-solving-using-computational-thinking/.

The NSF Computational Mechanics Vision Workshop

By | Events, Research

Over October 31 and November 1, 2019 MICDE hosted the 2019 Computational Mechanics Vision workshop that aimed to gather and synthesize future directions for computational mechanics research in the United States. Attended by more than 50 experts in various sub-disciplines of computational mechanics from across the country, including five National Science Foundation Program Directors, the group spent a day and a half brainstorming about the future of computational mechanics and defining new paradigms, methodologies and trends in this exciting and vast field. The workshop focused on four emerging areas in Computational Mechanics: Machine Learning, Additive Manufacturing, Computational Medicine, and Risk and Uncertainty Quantification. Operating through open discussions on talks by experts from within and beyond Computational Mechanics, and breakout sessions on the above four topics, the workshop participants arrived at a series of recommendations that could drive NSF’s investments in this field for the next decade and beyond.

To learn more about the event please visit micde.umich.edu/nsf-compmech-workshop-2019/.

46 Peta-FLOPS computation of defects in solid crystals is a finalist in the highest prize for scientific computing

By | HPC, News, Research

From left: Sambit Das, Phani Motamarri and Vikram Gavini

A team led by Prof. Vikram Gavini (Professor of Mechanical Engineering and MICDE affiliate) and including Dr. Sambit Das (MICDE Fellow) and Dr. Phani Motamarri (Assistant Research Scientist and MICDE affiliate), is one of two finalists nominated for this year’s Gordon Bell Prize. The award, generally considered to be the highest honor of its kind, worldwide, recognizes outstanding achievement in high-performance computing. Gavini’s team has developed a methodology that combines advanced finite-element discretization methods for Density Functional Theory (DFT)1 with efficient computational methodologies and mixed precision strategies to achieve a 46 Peta-FLOPS2 sustained performance on 3,800 GPU nodes of the Summit supercomputer. Their work titled “Fast, scalable and accurate finite-element based ab initio calculations using mixed precision computing: 46 PFLOPS simulation of a metallic dislocation3 system” also involved Dr. Bruno Turcksin and Dr. Ying Wai Li from Oak Ridge National Laboratory, and Los Alamos National Laboratory, and Mr. Brent Leback from NVIDIA Corporation.

Electron density contour of pyramidal II screw dislocation system in Mg with 61,640 electrons (6,164 Mg atoms).

First principle calculation methods4 have been immensely successful in predicting a variety of material properties.  These calculations are prohibitively expensive as the computational complexity scales with the number of electrons in the system. Prof. Gavini’s research work is focussed on developing fast and accurate algorithms for Kohn-Sham5 density functional theory, a workhorse of first principle approaches that occupies a significant fraction of the world’s supercomputing resources. In the current work, Dr. Das, Dr. Motamarri and Prof. Gavini used recent developments in the computational framework for real-space DFT calculations using higher-order adaptive finite elements, and pioneered algorithmic advances in the solution of the governing equations, along with a clever parallel implementation that reduced the data access costs and communication bottlenecks. This resulted in fast, accurate and scalable large-scale DFT calculations that are an order of magnitude faster than existing widely used DFT codes. They demonstrated an unprecedented sustained performance of 46 Peta-FLOPS on a dislocation system containing ~100,000 electrons, which is the subject of the Gordon Bell nomination.

Past winners of the Gordon Bell Prize have typically been large teams working on grand challenge problems in astrophysics, climate science, natural hazard modeling, quantum physics, materials science and public health. The purpose of the award is to track the progress over time of parallel computing, with particular emphasis on rewarding innovation in applying high-performance computing to applications in science, engineering, and large-scale data analytics. If you are attending the SuperComputing’19 conference this year in Denver, you can learn more about Dr. Das, Dr. Motamarri and Dr. Gavini’s achievement at the Gordon Bell Prize finalists’ presentations on Wednesday, November 20, 2019, at 4:15 pm in rooms 205-207

Related Publication: S. Das, P. Motamarri, V. Gavini, B. Turcksin, Y. W. Li, and B. Leback. “Fast, Scalable and Accurate Finite-Element Based Ab initio Calculations Using Mixed Precision Computing: 46 PFLOPS Simulation of a Metallic Dislocation System.” To appear in SC’19 Proceedings of the International Conference for High Performance Computing, Networking, Storage, and Analysis, Denver, CO, November 17–22, 2019.

[1] Density functional theory (DFT) is a computational quantum mechanical modeling method used in physics, chemistry and materials science to investigate the electronic structure (or nuclear structure) (principally the ground state) of many-body systems, in particular atoms, molecules, and the condensed phases. https://en.wikipedia.org/wiki/Density_functional_theory.
[2] A PETAFLOP is a unit of computing speed equal to one thousand million million (1015) floating-point operations per second.
[3] In materials science, dislocations are line defects that exist in crystalline solids.
[4] First principle calculation methods use the principle of quantum mechanics to compute properties directly from basic physical quantities such as, e.g., mass and charge.
[5] W. Kohn, L. J. Sham, Self-consistent equations including exchange and correlation effects, Phys. Rev. 140(4A) (1965) A1133.